I've not noticed a result like this before. Is this a valid result?

Message boards : Number crunching : I've not noticed a result like this before. Is this a valid result?
Message board moderation

To post messages, you must log in.

AuthorMessage
Dave Lewis

Send message
Joined: 12 Apr 99
Posts: 34
Credit: 53,432,603
RAC: 108
United States
Message 2029914 - Posted: 30 Jan 2020, 18:18:00 UTC
Last modified: 30 Jan 2020, 18:19:02 UTC

Whenever I see my computer produce an invalid result I try to check it in order to determine if I might have a hardware problem. This result had me scratching my head.

https://setiathome.berkeley.edu/workunit.php?wuid=3860988497

If you check the task details you'd notice that the validated result has the following for the stderr output:

<core_client_version>7.14.2</core_client_version>

while the stderr output for my computer (8817251) displays a more "normal" looking output, at least to my eyes and memory (which could be at fault) and is marked as invalid. Since there have been so many issues lately when I saw this I didn't know what to think or do other than post this question. Any feedback would be appreciated.
ID: 2029914 · Report as offensive
Profile Mr. Kevvy Crowdfunding Project Donor*Special Project $250 donor
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 15 May 99
Posts: 3778
Credit: 1,114,826,392
RAC: 3,319
Canada
Message 2029917 - Posted: 30 Jan 2020, 18:23:09 UTC - in response to Message 2029914.  

Please see the server issues thread. It appears that one-task validation is enabled.
ID: 2029917 · Report as offensive
Profile Schatten

Send message
Joined: 12 Oct 02
Posts: 18
Credit: 14,047,388
RAC: 9
Germany
Message 2029919 - Posted: 30 Jan 2020, 18:26:32 UTC - in response to Message 2029914.  

Looks very worrying to me. Like this: https://setiathome.berkeley.edu/workunit.php?wuid=3860615966 no Wingman needed anymore.

Normaly the task will be send out to a third computer and normally to need 2 not 1 with the same result.
ID: 2029919 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2029920 - Posted: 30 Jan 2020, 18:26:44 UTC - in response to Message 2029914.  

Yes, this is the result of adaptive replication being turned on and it improperly validated a bad result and tossed out your good result. This is a great example of what we were afraid of happening.

Without a proper stderr.txt for the validated result, can only guess what the results looked like. Based on the time taken for the task I assume it was processed as an early overflow.

However your result shows a normal search algorithm result of detected spikes, autocorrs, pulses, triplets and gaussians.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2029920 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14659
Credit: 200,643,578
RAC: 874
United Kingdom
Message 2029921 - Posted: 30 Jan 2020, 18:27:36 UTC - in response to Message 2029914.  

I think you're right to be worried. The 'wingmate' given the validation on this occasion has a lot of invalid tasks in their machine's history. The short runtime suggests that computer judged it to be an overflow (missing stderr output is not by itself a problem), whereas your computer ran to full term.

It's been confirmed that Eric is using some shortcuts to try and cut down the backlog of post-processing work (cut the queues to size), but this one looks like a shortcut too far.
ID: 2029921 · Report as offensive
Dave Lewis

Send message
Joined: 12 Apr 99
Posts: 34
Credit: 53,432,603
RAC: 108
United States
Message 2029923 - Posted: 30 Jan 2020, 18:30:42 UTC - in response to Message 2029917.  

Please see the server issues thread. It appears that one-task validation is enabled.


Thanks you Mr. Kevvy. I appreciate your quick answer. I'll read through that thread as it progresses in order to determine what this (one-task validation) means. My gut tells me that it isn't good since we've always depended on a wingman in the past to duplicate result - at least as far as I can remember.
ID: 2029923 · Report as offensive
Dave Lewis

Send message
Joined: 12 Apr 99
Posts: 34
Credit: 53,432,603
RAC: 108
United States
Message 2029924 - Posted: 30 Jan 2020, 18:35:01 UTC - in response to Message 2029921.  

I think you're right to be worried. The 'wingmate' given the validation on this occasion has a lot of invalid tasks in their machine's history. The short runtime suggests that computer judged it to be an overflow (missing stderr output is not by itself a problem), whereas your computer ran to full term.

It's been confirmed that Eric is using some shortcuts to try and cut down the backlog of post-processing work (cut the queues to size), but this one looks like a shortcut too far.


Thanks Richard. Due to low a serum caffeine level I hadn't gotten around to checking out the wingmate's validation record yet. Good catch. Yes, this is very worrisome.
ID: 2029924 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 2029925 - Posted: 30 Jan 2020, 18:36:10 UTC - in response to Message 2029923.  

The premise of adaptive replication is that it first determines hosts that have historically returned valid results in a timely manner consistently. This implies the hardware is being operated in a proper stable configuration. Then that host gets sent tasks with quorum of one.

This did not happen correctly on the host with a history of invalid results.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 2029925 · Report as offensive
Profile Schatten

Send message
Joined: 12 Oct 02
Posts: 18
Credit: 14,047,388
RAC: 9
Germany
Message 2029926 - Posted: 30 Jan 2020, 18:39:10 UTC - in response to Message 2029925.  

Also a nice example just got a invalid task on the CPU https://setiathome.berkeley.edu/workunit.php?wuid=3860616158
ID: 2029926 · Report as offensive
Dave Lewis

Send message
Joined: 12 Apr 99
Posts: 34
Credit: 53,432,603
RAC: 108
United States
Message 2029927 - Posted: 30 Jan 2020, 18:43:35 UTC - in response to Message 2029925.  

Thank you Keith for posting the link to the bad result in the most appropriate thread.
ID: 2029927 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14659
Credit: 200,643,578
RAC: 874
United Kingdom
Message 2029928 - Posted: 30 Jan 2020, 18:43:48 UTC

In both the examples in this thread, the first task to be reported to the server was given the validation. When the slower but more reliable reply was returned, it was different from the first - and got an immediate 'invalid', not an inconclusive / tiebreak outcome. This isn't the right solution to the over-full database.
ID: 2029928 · Report as offensive
Profile Schatten

Send message
Joined: 12 Oct 02
Posts: 18
Credit: 14,047,388
RAC: 9
Germany
Message 2029939 - Posted: 30 Jan 2020, 19:26:22 UTC - in response to Message 2029928.  

Let us hope that this king of Tasks will be resend again.
ID: 2029939 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14659
Credit: 200,643,578
RAC: 874
United Kingdom
Message 2029941 - Posted: 30 Jan 2020, 19:37:04 UTC

I have two workunits, both dated today, which follow the same pattern:

https://setiathome.berkeley.edu/workunit.php?wuid=3860489084
https://setiathome.berkeley.edu/workunit.php?wuid=3860488845

And they both feature the same wingmate - host 7941469 - as the opening post in this thread.
ID: 2029941 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 35466
Credit: 261,360,520
RAC: 489
Australia
Message 2029979 - Posted: 30 Jan 2020, 22:32:20 UTC

My 4th of these that have given an unreliable rig the nod. :-(

https://setiathome.berkeley.edu/workunit.php?wuid=3861228770

This idea was a really bad 1 for the science.

Cheers.
ID: 2029979 · Report as offensive
Tom*

Send message
Joined: 12 Aug 11
Posts: 127
Credit: 20,769,223
RAC: 9
United States
Message 2030028 - Posted: 31 Jan 2020, 4:21:29 UTC

https://setiathome.berkeley.edu/workunit.php?wuid=3861207454
ID: 2030028 · Report as offensive
Juhani Karjanlahti Project Donor
Volunteer tester

Send message
Joined: 23 Jan 03
Posts: 15
Credit: 83,675,733
RAC: 149
Finland
Message 2030030 - Posted: 31 Jan 2020, 4:40:19 UTC
Last modified: 31 Jan 2020, 4:42:07 UTC

I have exactly same question with this workunit
https://setiathome.berkeley.edu/workunit.php?wuid=3861231201

Why other machine producing massive amounts of invalid work with HD 5xxx is declared as correct result?
ID: 2030030 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 35466
Credit: 261,360,520
RAC: 489
Australia
Message 2030104 - Posted: 31 Jan 2020, 12:31:47 UTC

Thankfully I havn't had another of them for the last 12 hours.

Cheers.
ID: 2030104 · Report as offensive
Profile Mr. Kevvy Crowdfunding Project Donor*Special Project $250 donor
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 15 May 99
Posts: 3778
Credit: 1,114,826,392
RAC: 3,319
Canada
Message 2030129 - Posted: 31 Jan 2020, 17:26:44 UTC

Dr. Korpela has confirmed that normal two-quorum validation has been restored. Whew!
ID: 2030129 · Report as offensive
Kevin Olley

Send message
Joined: 3 Aug 99
Posts: 906
Credit: 261,085,289
RAC: 572
United Kingdom
Message 2030139 - Posted: 31 Jan 2020, 18:05:00 UTC - in response to Message 2030129.  

Dr. Korpela has confirmed that normal two-quorum validation has been restored. Whew!


Allow new tasks has now been set.
Kevin


ID: 2030139 · Report as offensive

Message boards : Number crunching : I've not noticed a result like this before. Is this a valid result?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.